Re: [RESEND PATCH net-next 1/1] net/mlx5: Clean up only new IRQ glue on request_irq() failure

[Mark Bloch <mbloch@nvidia.com>]

On 26/06/2025 9:04, Mohith Kumar Thummaluru wrote:
> The mlx5_irq_alloc() function can inadvertently free the entire rmap
> and end up in a crash[1] when the other threads tries to access this,
> when request_irq() fails due to exhausted IRQ vectors. This commit
> modifies the cleanup to remove only the specific IRQ mapping that was
> just added.
> 
> This prevents removal of other valid mappings and ensures precise
> cleanup of the failed IRQ allocation's associated glue object.
> 
> Note: This error is observed when both fwctl and rds configs are enabled.
> 

Please target net and not net-next.

Mark

> [1]
> mlx5_core 0000:05:00.0: Successfully registered panic handler for port 1
> mlx5_core 0000:05:00.0: mlx5_irq_alloc:293:(pid 66740): Failed to request irq. err = -28
> infiniband mlx5_0: mlx5_ib_test_wc:290:(pid 66740): Error -28 while trying to test write-combining support
> mlx5_core 0000:05:00.0: Successfully unregistered panic handler for port 1
> mlx5_core 0000:06:00.0: Successfully registered panic handler for port 1
> mlx5_core 0000:06:00.0: mlx5_irq_alloc:293:(pid 66740): Failed to request irq. err = -28
> infiniband mlx5_0: mlx5_ib_test_wc:290:(pid 66740): Error -28 while trying to test write-combining support
> mlx5_core 0000:06:00.0: Successfully unregistered panic handler for port 1
> mlx5_core 0000:03:00.0: mlx5_irq_alloc:293:(pid 28895): Failed to request irq. err = -28
> mlx5_core 0000:05:00.0: mlx5_irq_alloc:293:(pid 28895): Failed to request irq. err = -28
> general protection fault, probably for non-canonical address 0xe277a58fde16f291: 0000 [#1] SMP NOPTI
> 
> RIP: 0010:free_irq_cpu_rmap+0x23/0x7d
> Call Trace:
>   <TASK>
>   ? show_trace_log_lvl+0x1d6/0x2f9
>   ? show_trace_log_lvl+0x1d6/0x2f9
>   ? mlx5_irq_alloc.cold+0x5d/0xf3 [mlx5_core]
>   ? __die_body.cold+0x8/0xa
>   ? die_addr+0x39/0x53
>   ? exc_general_protection+0x1c4/0x3e9
>   ? dev_vprintk_emit+0x5f/0x90
>   ? asm_exc_general_protection+0x22/0x27
>   ? free_irq_cpu_rmap+0x23/0x7d
>   mlx5_irq_alloc.cold+0x5d/0xf3 [mlx5_core]
>   irq_pool_request_vector+0x7d/0x90 [mlx5_core]
>   mlx5_irq_request+0x2e/0xe0 [mlx5_core]
>   mlx5_irq_request_vector+0xad/0xf7 [mlx5_core]
>   comp_irq_request_pci+0x64/0xf0 [mlx5_core]
>   create_comp_eq+0x71/0x385 [mlx5_core]
>   ? mlx5e_open_xdpsq+0x11c/0x230 [mlx5_core]
>   mlx5_comp_eqn_get+0x72/0x90 [mlx5_core]
>   ? xas_load+0x8/0x91
>   mlx5_comp_irqn_get+0x40/0x90 [mlx5_core]
>   mlx5e_open_channel+0x7d/0x3c7 [mlx5_core]
>   mlx5e_open_channels+0xad/0x250 [mlx5_core]
>   mlx5e_open_locked+0x3e/0x110 [mlx5_core]
>   mlx5e_open+0x23/0x70 [mlx5_core]
>   __dev_open+0xf1/0x1a5
>   __dev_change_flags+0x1e1/0x249
>   dev_change_flags+0x21/0x5c
>   do_setlink+0x28b/0xcc4
>   ? __nla_parse+0x22/0x3d
>   ? inet6_validate_link_af+0x6b/0x108
>   ? cpumask_next+0x1f/0x35
>   ? __snmp6_fill_stats64.constprop.0+0x66/0x107
>   ? __nla_validate_parse+0x48/0x1e6
>   __rtnl_newlink+0x5ff/0xa57
>   ? kmem_cache_alloc_trace+0x164/0x2ce
>   rtnl_newlink+0x44/0x6e
>   rtnetlink_rcv_msg+0x2bb/0x362
>   ? __netlink_sendskb+0x4c/0x6c
>   ? netlink_unicast+0x28f/0x2ce
>   ? rtnl_calcit.isra.0+0x150/0x146
>   netlink_rcv_skb+0x5f/0x112
>   netlink_unicast+0x213/0x2ce
>   netlink_sendmsg+0x24f/0x4d9
>   __sock_sendmsg+0x65/0x6a
>   ____sys_sendmsg+0x28f/0x2c9
>   ? import_iovec+0x17/0x2b
>   ___sys_sendmsg+0x97/0xe0
>   __sys_sendmsg+0x81/0xd8
>   do_syscall_64+0x35/0x87
>   entry_SYSCALL_64_after_hwframe+0x6e/0x0
> RIP: 0033:0x7fc328603727
> Code: c3 66 90 41 54 41 89 d4 55 48 89 f5 53 89 fb 48 83 ec 10 e8 0b ed ff ff 44 89 e2 48 89 ee 89 df 41 89 c0 b8 2e 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 35 44 89 c7 48 89 44 24 08 e8 44 ed ff ff 48
> RSP: 002b:00007ffe8eb3f1a0 EFLAGS: 00000293 ORIG_RAX: 000000000000002e
> RAX: ffffffffffffffda RBX: 000000000000000d RCX: 00007fc328603727
> RDX: 0000000000000000 RSI: 00007ffe8eb3f1f0 RDI: 000000000000000d
> RBP: 00007ffe8eb3f1f0 R08: 0000000000000000 R09: 0000000000000000
> R10: 0000000000000000 R11: 0000000000000293 R12: 0000000000000000
> R13: 0000000000000000 R14: 00007ffe8eb3f3c8 R15: 00007ffe8eb3f3bc
>   </TASK>
> ---[ end trace f43ce73c3c2b13a2 ]---
> RIP: 0010:free_irq_cpu_rmap+0x23/0x7d
> Code: 0f 1f 80 00 00 00 00 48 85 ff 74 6b 55 48 89 fd 53 66 83 7f 06 00 74 24 31 db 48 8b 55 08 0f b7 c3 48 8b 04 c2 48 85 c0 74 09 <8b> 38 31 f6 e8 c4 0a b8 ff 83 c3 01 66 3b 5d 06 72 de b8 ff ff ff
> RSP: 0018:ff384881640eaca0 EFLAGS: 00010282
> RAX: e277a58fde16f291 RBX: 0000000000000000 RCX: 0000000000000000
> RDX: ff2335e2e20b3600 RSI: 0000000000000000 RDI: ff2335e2e20b3400
> RBP: ff2335e2e20b3400 R08: 0000000000000000 R09: 0000000000000000
> R10: 0000000000000000 R11: 00000000ffffffe4 R12: ff384881640ead88
> R13: ff2335c3760751e0 R14: ff2335e2e1672200 R15: ff2335c3760751f8
> FS:  00007fc32ac22480(0000) GS:ff2335e2d6e00000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00007f651ab54000 CR3: 00000029f1206003 CR4: 0000000000771ef0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> PKRU: 55555554
> Kernel panic - not syncing: Fatal exception
> Kernel Offset: 0x1dc00000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
> kvm-guest: disable async PF for cpu 0
> 
> 
> Fixes: 3354822cde5a ("net/mlx5: Use dynamic msix vectors allocation")
> Signed-off-by: Mohith Kumar Thummaluru<mohith.k.kumar.thummaluru@oracle.com>
> Tested-by: Mohith Kumar Thummaluru<mohith.k.kumar.thummaluru@oracle.com>
> Reviewed-by: Moshe Shemesh<moshe@nvidia.com>
> ---
>   drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c | 3 +--
>   1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
> index 40024cfa3099..822e92ed2d45 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
> @@ -325,8 +325,7 @@ struct mlx5_irq *mlx5_irq_alloc(struct mlx5_irq_pool *pool, int i,
>   err_req_irq:
>   #ifdef CONFIG_RFS_ACCEL
>       if (i && rmap && *rmap) {
> -        free_irq_cpu_rmap(*rmap);
> -        *rmap = NULL;
> +        irq_cpu_rmap_remove(*rmap, irq->map.virq);
>       }
>   err_irq_rmap:
>   #endif